S. Annas, R. Arisandi
There are many methods that can be used in forecasting rainfall occurrences, ranging from classical statistics method to the neural network-based method. This study applied both methods of a multivariate transfer function and an artificial resilient backpropagation neural network (RProp) for forecasting rainfall in Makassar area, South Sulawesi of Indonesia. First, the multivariate transfer function was used in forecasting the amount of rainfall that caused by several variables such as air humidity, air temperature, air pressure, and wind velocity. This method obtained a result of forecasting simultaneously and able to provide an information on variables that have a significant effect to the rainfall occurance. However, it cannot give a forecasting result with maximum accuracy. To overcome this problem, a method of Rpro was then used. This proposed method obtained the error average percentage of the difference between the actual data and the prediction data is 0.197%. The combianation of both methods produced several significant variables that affect the amount of rainfall such as air temperatur, air pressure, and velocity wind and also reslted a maximum accuracy for forcasting to rainfall occurance. © 2017 Author(s).
Departement of Statistics, Universitas Negeri Makassar, Kampus UNM Parangtambung, Makassar, 90223, Indonesia
